Three-State Conical Intersections in Nucleic Acid Bases

View Sections. ACS2GO © 2018. ← → → ←. loading. To add this web app to the home screen open the browser option menu and tap on Add to hom...
0 downloads 0 Views 269KB Size
7538

J. Phys. Chem. A 2005, 109, 7538-7545

Three-State Conical Intersections in Nucleic Acid Bases Spiridoula Matsika* Department of Chemistry, Temple UniVersity, Philadelphia, PA, 19122 ReceiVed: March 15, 2005; In Final Form: June 2, 2005

The involvement of three-state conical intersections in the photophysics and radiationless decay processes of the nucleobases has been investigated using multireference configuration interaction methods. Three-state conical intersections have been located for the pyrimidine base, uracil, and the purine base, adenine. In uracil, a three-state degeneracy between the S0, S1, and S2 states has been located at 6.2 eV above the ground-state minimum energy. This energy is 0.4 eV higher than vertical excitation to S2 and at least 1.3 eV higher than the two-state conical intersections found previously. In adenine, two different three-state degeneracies between the S1, S2, and S3 states have been located at energies close to the vertical excitation energies. The energetics of these three-state conical intersections suggest they can play a role in a radiationless decay pathway present in adenine. The existence of two different seams of three-state conical intersections indicates that these features are common and complicate the potential energy surfaces of adenine and possibly many other aromatic molecules.

1. Introduction Although the importance of two-state conical intersections in nonadiabatic processes has been established,1,2 the occurrence and relevance of accidental three-state conical intersections is only now emerging. Three-state degeneracies imposed by symmetry had been studied in the context of the Jahn-Teller problem for many decades,3 but only minor attention had been given to accidental three-state degeneracies in molecules.4 Because most molecular systems in nature have low or no symmetry, these accidental intersections may have a great impact on the photophysics and photochemistry of molecular systems, as has been found in accidental two-state intersections.1,2,5,6 Three-state degeneracies can provide a more efficient relaxation pathway when more than one interstate transition is needed, they introduce more complicated geometric phase effects,7-9 and they can affect the dynamics and available pathways.10 The exact way these features affect the dynamical behavior of molecules is a question that has yet to be studied extensively. According to the noncrossing rule,11 three states can become degenerate in a subspace of dimension Nint - 5, where Nint is the number of internal coordinates. For this reason, they cannot be found in molecular systems with less than five internal coordinates. Even in systems with enough degrees of freedom, the dimensionality makes their location difficult. Until recently, the perception was that three-state degeneracies are extremely rare, mostly because of the inability to locate them. Recent developments of efficient algorithms, however, have made these calculations possible and have revealed that three-state conical intersections exist in many organic radicals12-14 and may play an important role on their spectroscopic or photophysical properties. Both excited-state dynamics and ground-state vibrational spectra can be affected by these features. More recent studies indicate the existence of three-state conical intersections in closed-shell systems also.10,15 Systems with high density of electronic states are more likely to exhibit three-state crossings if they have enough degrees of * Corresponding author. E-mail address: [email protected].

freedom. Examples of systems with many, closely spaced, excited electronic states are radicals with closely spaced Rydberg excited states, such as the ethyl and allyl radical, where the 3s and 3p Rydberg states can intersect.12,13 Another very important category of molecules that have many electronic states close in energy are conjugated systems. An example of a conjugated open-shell system is the pyrazolyl radical, which has been found to have three-state conical intersections only 3000 cm-1 above the ground-state minimum.14 In this case, the ground state is one of the degenerate states, and complicated vibronic spectra are expected, and have been observed experimentally.16 In a closed shell aromatic system, excitations to π* orbitals create a manifold of excited states. Some of the most important molecules in nature, such as the nucleobases discussed here, are closed-shell aromatic molecules. The excited-state dynamics of nucleic acid bases has gained much attention recently.17 The problem is important because of the biological significance of the bases. These bases are the dominant chromophores in nucleic acids, and their photochemical and photophysical properties are implicated in the behavior of the nucleic acids upon irradiation.18-20 The excited-state lifetimes have been reported recently using many different techniques.21-39 These studies report lifetimes on the order of femtoseconds and suggest that nonradiative relaxation proceeds on an ultrafast time scale to the ground state, with the extra energy being transformed into heat. The involvement of two-state conical intersections in the nucleobases has been examined by several groups. Conical intersections were found in cytosine,40,41 uracil,42 and adenine,43,44 and pathways were suggested that could explain mechanisms for radiationless decay to the ground state. A recent work examined three-state near degeneracies in cytosine.15 A near-degeneracy point was found ca. 1.5 eV higher than the vertical excitation to the bright state of cytosine using perturbative methods (CASPT2). At the complete-active space selfconsistent-field (CASSCF) level this point is ca. 2 eV lower and close to vertical excitations. These discrepancies in the energetics are probably due to the fact that an algorithm to locate

10.1021/jp0513622 CCC: $30.25 © 2005 American Chemical Society Published on Web 07/29/2005

Three-State Conical Intersections in Nucleic Acid Bases three-state conical intersections was not used, so these points are not optimized points on the seam. Although the authors propose that the three-state intersection is important in the radiationless decay, the energetics at the CASPT2 level suggest the opposite. The purpose of this work is to examine the different possibilities in which three-state conical intersections may appear in the potential energy surfaces of the nucleobases and to investigate in which cases they are expected to be important. Three-state conical intersections in two nucleobases, uracil and adenine, are reported using multireference configuration interaction (MRCI) methods. Section 2 discusses the theoretical methods used for obtaining the electronic energies and gradients and locating conical intersections. Three-state conical intersections have been located between the S0, S1, and S2 states in uracil and between the S1, S2, and S3 states in adenine. The most stable tautomer of adenine, 9H-adenine, is examined. These intersections are presented and discussed in Section 3, and conclusions are given in Section 4. 2. Methods The adiabatic energies and wave functions were determined at the MRCI level using orbitals from a state-averaged CASSCF procedure. The double-ζ plus polarization (cc-pvdz) Gaussian basis sets of Dunning were used for all atoms.45 Optimizations were carried out using first-order MRCI expansions. Singlepoint MRCI calculations, with expansions, including single and double excitations, were used to refine the energies. The algorithm for locating three-state conical intersections has been described previously.12 A modified version of the COLUMBUS suite of programs46 was used, where the algorithms for locating two- and three-state conical intersections12,47,48 have been included. The algorithms use analytic gradients for MRCI wave functions available in COLUMBUS.49 The specific details of the multireference calculations for uracil and adenine are given below. 2.1. Uracil. The justification and details of the choice of the MRCI expansions used in uracil are given in our previous work.42 Here, a summary of these methods is provided. A stateaveraged CASSCF was used, where the ground state and two excited singlet electronic states were included in the average. An active space with 12 electrons in 9 orbitals, denoted (12, 9), was used in both CASSCF and MRCI in the optimization calculations. The designation (n, m) is used here for an active space of n electrons in m orbitals. It was found previously that this active space, where the nine active orbitals were the eight π and one of the two nO valence orbitals, gave energies very similar to the ones obtained from an active space that included all valence orbitals, eight π and both nO orbitals. Two different MRCI expansions were used for the optimizations. In the first expansion, all σ electrons were frozen, and single excitations only were allowed out of the active space. The number of configuration state functions (CSFs) was 607 320, and this expansion is denoted here as MRCI1. In the second expansion, only the core 1s electrons were frozen, and single excitations were allowed out of the σ orbitals as well. This expansion consisted of 9 518 670 CSFs and is denoted MRCI1σ. MRCI1σ was used to reoptimize the three-state conical intersection because the inclusion of dynamical correlation changed the energies of degenerate states so that they became separated by more than 0.4 eV. This observation will be discussed further in Section 3. The energies were refined using an MRCI expansion with a (14, 10) reference space, which included σπ correlation (denoted MRCIσπ) as described in the previous work.42 This MRCIσπ expansion consisted of ca. 100 million CSFs.

J. Phys. Chem. A, Vol. 109, No. 33, 2005 7539

Figure 1. Molecular structure and atom numbering for (a) uracil and (b) adenine.

2.2. Adenine. Adenine has 10 π and 3 lone-pair nN orbitals from the nitrogen atoms, giving 13 active orbitals with 18 electrons. An active space (18, 13) generates too many references in either the MCSCF or the MRCI procedures. However, the orbitals localized mainly on the five-membered ring do not participate in the first few excited states that we are interested in; rather it is the orbitals localized on the six-membered ring that are important for these states. Initial trials showed that a (12, 9) active space, where the nine orbitals were seven π and two nN orbitals, was adequate for the lowest excited states. The orbitals were obtained from a state-averaged CASSCF using the (12, 9) active space and averaging over five singlet electronic states. MRCI expansions with references generated from this active space were used for all gradient calculations. As in uracil, two different expansions were generated. In the first one (denoted MRCI1), all σ electrons were frozen, and single excitations from the (12, 9) active space generated 770 616 CSFs. In the second (denoted MRCI1σ), only the 1s core electrons were frozen, and single excitations from the σ orbitals were allowed as well, giving 15 096 312 CSFs. MRCI1σ was used in adenine for the same reason it was used in uracil, to reoptimize a conical intersection when correlation changed the energy separation of the states by more than 0.4 eV. Refinement of the energetics was obtained by using extended MRCI expansions. All single and double excitations from the (12, 9) active space and only single excitations from all σ orbitals were included. The 10 1s core orbitals and the 10 highest in energy virtual orbitals were kept frozen. The resulting expansion is denoted MRCIσπ and includes ca. 200 million CSFs. In Cs symmetry, the resulting expansions for A′ and A′′ symmetries are 100 880 026 and 101 068 778 CSFs, respectively. When the geometry is almost coplanar, Cs symmetry is employed for single-point energy calculations using MRCIσπ because the computational cost is significantly reduced. 3. Results and Discussion The structure and conventional labeling of the atoms for uracil and adenine (9H-adenine) are shown in Figure 1. Below, we discuss the three-state conical intersections found in these molecules. The designation of conical intersections between two states, SI, SJ, is ciIJ and between three states, SI, SJ, SK, is ciIJK. Geometries at a conical intersection ciIJ are given as Rx(ciIJ), and geometries of minimum energy points on a surface SI as Re(SI). 3.1. Uracil. In a previous work, we located conical intersections in uracil between the S2 and S1 surfaces and between the S1 and the ground state, S0.42 The two-state conical intersections in uracil facilitate efficient radiationless decay from the bright S2 state to the ground state. Because a radiationless decay involves the three surfaces S0, S1, and S2, the existence of a three-state conical intersection between these states could expedite the overall process if the seam is energetically accessible. Here, a three-state conical intersection between the three states participating in this mechanism has been located.

7540 J. Phys. Chem. A, Vol. 109, No. 33, 2005

Matsika

TABLE 1: Energies in eV for the Three Lowest States of Uracil at the Equilibrium of the Ground State Re(S0) (vertical excitations) and at Conical Intersections Rx(ciIJ) or Rx(ciIJK)a Re(S0)b

Rx(ci12)b

Rx(ci01)b

Rx(ci012)

2.15 (1.87) 4.47 (3.96) 6.83 (5.90) S0 0 S1 5.44 (4.80)* 5.37 (4.83)* 4.47 (4.29) 6.83 (6.37)* S2 6.24 (5.79) 5.37 (4.97) 7.62* 6.83 (6.56)

Rx2(ci012) (6.04) (6.20)* (6.25)

a The energies in bold in each column correspond to the optimized degenerate states I, J or I, J, K. MRCI1 results are shown (the zero is set to -412.624433 au), and in parentheses are the single-point energies obtained using MRCIσπ (the zero is set to -412.815540 au). States that have nπ* character are designated with an * as superscript. b Taken from ref 42.

Table 1 gives the energy for the three-state conical intersection ci012, as well as the previous results of vertical excitation energies and two-state conical intersections. At the MRCI1 level, the minimum energy point on the seam, Rx(ci012), is 6.83 eV above the ground-state minimum. This energy is 0.6 eV higher than that of the vertical excitation to the bright state S2 and 1.5 and 2.4 eV higher than that of the two-state conical intersections ci12 and ci01, respectively (at the MRCI1 level). The results for two-state conical intersections are taken from our previous work.42 When MRCIσπ is used to recalculate the energies of S0, S1, and S2 at the conical intersection Rx(ci012), they are found to be separated by 0.7 eV. This is because dynamical correlation affects the energy of the three states to different degrees. To improve the energy separation at the more correlated level, the three-state conical intersection is located using the MRCI1σ expansion, which includes single excitations from the σ orbitals. At this new geometry, which is denoted Rx2(ci012), the energies of the three states at the MRCIσπ level are separated by 0.2 eV, and the average energy is 6.16 eV. The vertical excitation energy to the bright S2 state is 5.8 eV using MRCIσπ, so the three-state degeneracy is 0.4 eV higher. The energy of the twostate conical intersections are again 1.3 and 2 eV lower than that of the three-state degeneracy. In summary, the ci012 conical intersection is energetically too high at every level of theory investigated, and this makes it less likely to directly participate in a radiationless decay mechanism. The geometry at ci012 is shown in Figure 2 in comparison with the geometries at the minimum of the ground state and at the other two-state conical intersections, ci01 and ci12, from our previous work. A more complete list of internal coordinates is given as Supporting Information. The geometry at ci012 resembles the geometry at ci01, but the molecule is even further distorted. The most important distortion involves the oxygen O8 atom. The C4-O8 bond distance increases by more than 0.1 Å, and O8 is moved away from planarity. The dihedral angle ∠O8C4N3H3, which indicates the displacement of O8 away from planarity, is between 0 and 6° in all other geometries, but increases to 39° at ci012. H5 is still almost perpendicular to the original plane of the ring, as shown in Figure 2. The dihedral angle ∠H6C6C5H5 is 108° for the structures of both ci01 and ci012 conical intersections. Rx2(ci012) differs from Rx(ci012) mainly in the bond distances, which are shown in parentheses in Figure 2, but the main characteristics and dihedral distortions of the structures are similar. The three-state conical intersection occurs when the two different seams ci01 and ci12 cross. When the C5-C6 bond stretches from 1.35 to 1.48 Å, the S2 state (ππ*) is stabilized, while the S1 (nOπ*) is not affected, and the two states become degenerate, forming ci12. Breaking of the planarity of the ring

Figure 2. Geometry of uracil at (a) the S0 minimum, (b) the S1-S2 conical intersection, (c) the S0-S1 conical intersection, and (d) the threestate S0-S1-S2 conical intersection. All bond distances between nonhydrogen atoms are shown in Å. MRCI1 results are shown, except in (d), where the MRCIσ results are also shown in parentheses. More internal coordinates are given as Supporting Information.

by twisting of the C5-C6 bond and pyramidalization of C5 destabilizes the ground state by more than 4 eV, causing it to meet the ππ* state, forming the ci01 conical intersection. This distortion greatly destabilizes the nOπ* state which is 3 eV higher at the ci01 point. To bring together all three states, the geometry has to change in a way that the nOπ* state will be stabilized and the seam of ci01 will rise in energy to meet the nOπ*. This happens when the C4-O8 bond is extended, while the nonplanar distortions remain. 3.2. Adenine. Adenine, like the other nucleobases, has ultrafast excited-state lifetimes and is expected to decay nonradiatively to the ground state, possibly through conical intersections of an excited state with the ground state. Indeed, theoretical work has shown the existence of conical intersections of ππ*, nπ*, or πσ* states with the ground state, and radiationless decay has been proposed through these conical intersections.43,44 Here, we only focus on the excited states and the conical intersections between them. As will be seen in detail in the following section, the excited-states manifold for adenine is more complicated than that for uracil, and careful consideration of the interaction of excited states is important in order to understand its photodynamics. The closely spaced excited states suggest the existence of three-state conical intersections, and investigating this possibility is the main object of this work. Because adenine has not been studied previously using MRCI methods, before presenting the three-state conical intersections, the vertical excitation energies are presented and discussed. This is the starting point to determine which states are relevant and should be considered further. 3.2.1 Vertical Excitations. To calculate vertical excitation energies, the equilibrium geometry of adenine in the ground state is obtained by optimization at the MRCI1 level. The molecule is almost planar, but the amino group is pyramidalized (∠H10aN10C6N1 ) 15°, ∠H10bN10C6C5 ) -14°), as was found in previous theoretical studies.44,50-55 The geometry is shown in Figure 3 and is in good agreement with previous theoretical results. This figure includes only bond lengths, but a more complete list of internal coordinates is given as Supporting Information. Although the symmetry at equilibrium is deviated from Cs, it is computationally very advantageous to be able to employ symmetry in the calculations. For this reason, the effect of the nonplanar distortion on the vertical excitation energies

Three-State Conical Intersections in Nucleic Acid Bases

Figure 3. Geometry of adenine at (a) the S0 minimum, (b) the threestate S1-S2-S3 (ππ*, nπ*, nπ*) conical intersection, (c) the threestate S1-S2-S3 (ππ*,ππ*, nπ*) conical intersection, and (d) the S1S2 conical intersection. All bond distances between non-hydrogen atoms are shown in Å. MRCI1 results are shown, except in (c), where the MRCIσ results are also shown in parentheses. More internal coordinates are given as Supporting Information.

TABLE 2: Energies in eV for the Five Lowest States of Adenine at the Equilibrium Geometry of the Ground State Re(S0) (vertical excitations) and at Conical Intersections Rx(ciIJ) or Rx(ciIJK)a S0 S1 S2 S3 S4

Re(S0)

Rx(ci123)

Rx(ci123′)

Rx2(ci123′)

Rx(ci12)

0 5.68 (5.67) 5.85 (5.68)* 6.55 (6.11) 6.30 (6.25)*

0.43 (0.28) 5.72 (5.48)* 5.72 (5.62) 5.72 (5.67)* 6.91 (6.33)

0.85 (0.75) 6.24 (5.78) 6.24 (5.84)* 6.24 (6.18) 7.72 (7.42)*

(0.73) (5.86) (6.06)* (6.06) (7.37)*

0.27 (0.20) 5.52 (5.24)* 5.52 (5.41) 6.44 (6.07) 6.31 (6.17)*

a The energies in bold in each column correspond to the optimized degenerate states I, J or I, J, K. MRCI1 results are shown (the zero is set to -464.68301 au), and in parentheses are the single-point energies obtained using MRCIσπ (the zero is set to -464.923845 au). States that have nπ* character are designated with an * as superscript.

has been investigated. The vertical excitation energies were calculated first using no symmetry and then using Cs symmetry. It was found that the effect is small, in the range 0.01-0.05 eV, allowing Cs symmetry to be used with the high-level, computationally expensive calculations. Vertical excitation energies for adenine are given in Table 2. In this table, the common designation of excited states as S1, S2, etc., given by their energetic ordering, has been used. The character of the states is specified by using an * for the nπ* (A′′) states. At different points of the configurational space, this ordering changes. The energies at the MRCI1 and MRCIσπ levels are listed. The excitation energies and even the ordering of the states are affected by the inclusion of dynamical correlation. The first two excited states are nπ* and ππ*, and they are practically degenerate at the MRCIσπ level, with the ππ* slightly lower. The most dramatic effect of dynamical correlation is on the third excited-state S3, which is a second ππ* state. This state is at 6.55 eV at the MRCI1 level and at 6.11 eV using MRCIσπ, and consequently, the energy separation between the two ππ* states changes from 0.9 to 0.4 eV as the theory improves. The fourth excited state is a second nπ* state at 6.25 eV above the ground state at the MRCIσπ level. In summary, there are four excited states, two ππ* and two nπ* states, at excitation energies between 5.5 and 6.3 eV. The ordering predicted at the MRCIσπ level is ππ*, nπ*, ππ*, nπ*.

J. Phys. Chem. A, Vol. 109, No. 33, 2005 7541 The excited states of adenine have been examined theoretically before using various methods, such as perturbation theory,44,54,56 TDDFT,43,52 and CIS.55 Vertical excitations using TDDFT52 gave energies 4.97, 5.08, and 5.35 eV for nπ*, ππ*, and ππ* states, respectively. Earlier, CASPT256 vertical excitation energies were calculated to be 5.13, 5.20, 6.15, and 6.86 eV, while in a very recent work,44 CASPT2 calculations give 4.85, 4.90, 5.50, and 5.68 eV. The ordering of the states in both cases is ππ*, ππ*, nπ*, nπ*. All calculations agree that the four lowest states are two nπ* and two ππ* states, but their relative ordering varies between different theoretical methods. In all cases, the two ππ* states are close in energy. The nπ* state, however, is predicted very close to the first ππ* state by TDDFT and our MRCI results, but much higher than the ππ* states when CASPT2 is used. It is obvious that the excitation energies and ordering of the states are very sensitive to the level of theory and correlation, and accurate benchmarks are not available for a definite comparison. MRCI methods are superior to TDDFT and CASPT2 when all electrons are correlated and the size of basis set is adequate. Here, it is not possible to fulfill both of these requirements, and so the best possible description must be compromised. Experimental adiabatic energies have been obtained in the gas phase. Recent resonant two-photon ionization experiments37 predict structured vibronic spectra where the nπ* is the lowest excited state adiabatically at ca. 4.4 eV, but the ππ* state is only 600 cm-1 higher at 4.47 eV. Nir et al.30 and Lu¨hrs et al.32 report a 0-0 band at 4.47 eV. This ordering, however, corresponds to adiabatic energies and cannot be compared directly with the vertical excitations because the ordering may change. The UV-vis absorption spectrum of adenine has a lowenergy band with a peak around 4.9 eV in the gas phase and at 4.8 eV in aqueous solution.57 The vertical excitation energies obtained in this work are higher than both the other theoretical results and the experimental absorption maximum, although in complicated spectra, like here, it is not always true that the absorption maximum corresponds to the vertical excitation. Regardless of the above uncertainties, the results presented here, namely the presence and location of three-state conical intersections between the excited states, will not be affected qualitatively by changes in these energies. This claim is based on two facts. First, conical intersections were optimized using different MRCI expansions, and the additional correlation did not remove the degeneracies, but rather displaced them moderately. Second, the three-state conical intersections found exist independently of the ordering of states at vertical excitations. 3.2.2. ππ* - nπ* - nπ* Conical Intersection. Starting at the equilibrium geometry, where the first three excited states have character ππ*, nπ*, nπ* at the MRCI1 level, a conical intersection between these states was located. This conical intersection will be denoted ci123 and has energy 5.72 eV above the ground-state minimum at the MRCI1 level. At this level, the vertical excitation energies are 5.66, 5.85, and 6.27 eV, so the three-state conical intersection has about the same energy as vertical excitation to S1. The fourth excited state (ππ*) is ca. 1 eV above the three-state conical intersection. Using MRCIσπ, the three energies are separated by less than 0.2 eV and have an average of 5.59 eV. This energy is lower than the vertical excitation energy to S1 at the MRCIσπ level, which is 5.67 eV. It is important to note that the intersecting states correlate to the S1, S2, S4 states at vertical excitation but have become the S1, S2, S3 states at this geometry, with the fourth state being at 6.33 eV. An S3-S4 conical intersection is probably the cause of this switching of states.

7542 J. Phys. Chem. A, Vol. 109, No. 33, 2005

Figure 4. Five vectors defining the branching space for the ci123 conical intersection in adenine. The three top vectors correspond to the coupling vectors and the two bottom to the energy difference gradient vectors.

The geometry at ci123 is almost planar and is shown in Figure 3. Conical intersections involving the ground state, as the one found in uracil, require substantial geometrical distortions because, initially, the energy difference between the ground state and the excited state is more than 4 eV. Conical intersections, however, involving only excited states that are close energetically even at vertical excitation, require much smaller distortions. The distortions at ci123 involve mainly the bond distances in the six-membered ring because the states involved are described by orbitals mainly localized in this ring. The largest bond length distortions occur on C6-N1, which increases from 1.33 to 1.40 Å and on C4-C5, which increases from 1.38 to 1.42 Å. Excitation from the lone pair n to a π* orbital weakens the C6N1 double bond, and the distortions observed facilitate stabilization of the second nπ* state so that it lowers its energy to meet S1 and S2. The branching space for this three-state conical intersection is spanned by five vectors, instead of the usual two vectors in a two-state conical intersection.6,58 Nuclear motion along these five vectors will remove the degeneracy of the states linearly. At the point of degeneracy, the vectors are not unique because the degenerate eigenfunctions can mix in any way. For two vectors, one can define an orthogonal transformation of the eigenfunctions to obtain the orthogonal unique vectors,6 but this is not possible for the five vectors. In Figure 4, the five vectors are shown taken from a point slightly away from the degeneracy, so that the mixing of eigenvectors can be avoided. The visualization program MOLDEN59 was used to generate the pictures of the vectors. The vectors involve mainly distortions of the six-membered ring, either in-plane bond distortions and bending, or out-of-plane ring-puckering motion. Because the symmetry is almost Cs, and the states crossing are A′(ππ*), A′′(nπ*), and A′′(nπ*), the five vectors have symmetries 3a′ + 2a′′. There are two vectors defined by the energy difference gradients of two pairs of states, and these are always totally symmetric (a′). The other three vectors correspond to the three couplings between the states, giving two a′′ couplings and one a′. The a′′ couplings correspond to coupling between each of the nπ* states with the ππ* state. 3.2.3. nπ* - ππ* - ππ* Conical Intersection. A second seam of three-state conical intersections between the S1, S2, and S3 states has been located in which the character of the states is nπ*, ππ*, ππ*. This conical intersection is denoted ci123′,

Matsika

Figure 5. Five vectors defining the branching space for the ci123′ conical intersection in adenine. The three top vectors correspond to the coupling vectors and the two bottom to the energy difference gradient vectors.

and the energies are given in Table 2. The energy of the conical intersection is 6.24 eV at MRCI1, which is higher than ci123 but still in the range of vertical excitation energies. The fourth excited state, which is an nπ* state, is more than 1.5 eV higher in energy. At the MRCIσπ level, the energies are separated by 0.4 eV, with an average of 5.93 eV. To reduce the separation, this conical intersection was reoptimized using the MRCI1σ expansion in a similar way to that in uracil. At this new geometry, Rx2(ci123′), the energies are separated by 0.2 eV with an average of 5.99 eV. The geometry at the ci123′ conical intersection is shown in Figure 3. Like ci123, the structure is planar, but the ci123′ crossing involves larger distortions than those of the ci123 crossing, with the main changes involving bond distances in the six-membered ring. At this geometry, the ππ* states come closer in energy, and the second nπ* state is destabilized. The distortions that occur facilitate stabilization of the second ππ* state and destabilization of the first ππ* and the nπ* states. The C2-N3, C4-C,5 and C5-C6 bond lengths increase by 0.12, 0.08, and 0.09 Å, respectively, in agreement with the fact that a ππ* state is stabilized by relaxing the double bonds. The angles C4-N3-C2 and the angles of N10 with the ring also change. In Figure 5, the five vectors spanning the branching space are shown. Here, again, the symmetries of the vectors are 3a′ + 2a′′, as can be seen in the figure. The vectors involve mainly distortions of the six-membered ring as was seen in ci123, although here, the N9 atom is involved in one of the out-ofplane modes. 3.2.4. S1-S2 Conical Intersection Seam. Each of the threestate conical intersection seams has dimension Nint - 5 ) 39 - 5 ) 34. In the other five degrees of freedom, the degeneracy is lifted linearly. In three-dimensional subspaces of this fivedimensional branching space, the degeneracy can be lifted partially, and two of the three states will remain degenerate.7,8,13 These two-state conical intersection seams originating from the three-state conical intersection have been studied in the allyl radical.13 In adenine, different seams of two-state conical intersections originate from each of the three-state conical intersections, leading to a great number of two-state conical intersections at energies lower than the three-state seams. This picture will produce very complicated excited-state potential energy surfaces, and the photodynamics in these surfaces is expected to be affected.

Three-State Conical Intersections in Nucleic Acid Bases

J. Phys. Chem. A, Vol. 109, No. 33, 2005 7543

Figure 6. Minimum energy pathway (a) from the ci123 conical intersection to ci12 and (b) from the ci123′ conical intersection to ci12

Figure 8. Diagram of the energy levels at the main points discussed in this work using MRCIσπ; (a) the S0, S1, S2 states of uracil and (b) S0, S1, S2, S3, S4 states of adenine.

Figure 7. Two vectors defining the branching plane for the ci12 conical intersection in adenine

One of these two-state conical intersection seams is explored here. Starting from the three-state conical intersection ci123 (Figure 6a) or ci123′ (Figure 6b) the S1-S2 two-state conical intersection minimum was optimized. Because the initial point was already on the seam, the search moved along the seam until it found the minimum-energy point, ci12. The resulting paths are energy-minimized sections of the seam connecting each of the three-state conical intersections, with the minimum-energy point of the two-state conical intersection ci12. Figure 6 shows the energies of the ground and four excited states as a function of the bond distance R(C2-N3). Although all coordinates change during the optimization (there are no geometrical constraints imposed), the bond distance R(C2-N3) is chosen to be plotted because it changes smoothly along both paths. It increases from 1.33 to 1.39 Å along the path from ci123 to ci12 and decreases from 1.45 to 1.39 Å along the path from ci123′ to ci12. The energy of the resulting two-state conical intersection is 5.5 eV at the MRCI1 level and 5.3 eV at the MRCIσπ (see Table 2) and is lower than all excitation energies. An S1-S2 conical intersection involving ππ* and nπ* states has been reported recently by Perun et al.44 Figure 7 shows the two vectors spanning the branching plane for ci12 at the minimum-energy point of the seam. The coupling mode a′′ corresponds to the coupling between the ππ* and nπ* states and resembles outof-plane coupling modes of the ci123 and ci123′ three-state conical intersections. This is expected because ci123 and ci123′ include this pair of states, and its coupling should be part of their five branching vectors. This a′′ mode involves mainly motion of the N1-C2-N3 moiety. The a′ mode involves inplane distortions of the six-membered ring. 3.3. Effect of Electron Correlation on the Location of Conical Intersections. When locating conical intersections in systems of the size of uracil and adenine, one is forced to use a method that does not include all the required correlation and is less accurate than the method used to calculate energies at single points on the potential energy surface. For this reason, the energy differences obtained at the lower level of theory (here

MRCI1) are different from the ones at the higher level (MRCIσπ), and the predicted points of conical intersections are shifted. Furthermore, different points on the surface and different states gain different correlation energy. For example, the energy of the S1 state in adenine is not affected by dynamical correlation at Re(S0) but is lowered by 0.46 eV at the distorted geometry Rx(ci123′). How close the conical intersection predicted at the lower level of theory is to the one at a higher level of theory depends on the particular case and varies between different crossings, even within the same molecule. In this study, the ci123 conical intersection in adenine is located more accurately than that of the ci123′ at the MRCI1 level. The separation of energies at the MRCIσπ level is an indication of the quality of prediction of the conical intersection at the MRCI1 level. The ci012 conical intersection in uracil is also poorly described at MRCI1. These energies are given in Tables 1 and 2 and are discussed in detail in the previous sections. In this work, a procedure was used in which an intermediate expansion was used to refine the conical intersection if the energy difference of the states that should be degenerate at the MRCIσπ level was more than 0.3 eV. Optimization could not be done at the MRCIσπ level because the size of this expansion was more than 100 million CSFs. So, whenever the small expansion was inadequate, an intermediate expansion was used to locate the conical intersections. The choice of the particular intermediate expansion MRCIσ was based on the fact that single excitations from the σ orbitals were found to be important for more accurately calculating the excitation energies in uracil.42 Qualitatively, the conical intersection does not change much between the different expansions in general. In uracil, the reoptimization of the three-state conical intersection changed the energy difference between the states from 0.6 to 0.2 eV. The average of the three energies changed from 6.19 to 6.16 eV. The bond lengths changed from 0.01 to 0.06 Å, and the angles changed from 0 to 18°. The 18° occurred in one dihedral angle, while most of the other angles had very small changes. The energy difference in adenine changed from 0.4 to 0.2 eV. The average of the energies of the three states changed from 5.93 to 5.99 eV. The average bond distance change was 0.02 Å, with the largest being 0.07 Å, and the largest angle change was 3°. 4. Conclusions In this work, the existence of three-state conical intersections in the nucleobases uracil and adenine has been investigated. Figure 8 summarizes the energetics of the points of interest for both molecules and aims at emphasizing their differences. A

7544 J. Phys. Chem. A, Vol. 109, No. 33, 2005 three-state conical intersection between the S0, S1, and S2 states has been located in uracil at an energy 6.2 eV above the groundstate minimum energy. This energy is ca. 0.4 eV higher than vertical excitation to S2, 1.3 eV higher than the S1-S2 twostate conical intersection, and 2 eV higher than the S1-S0 conical intersection found previously. Consequently, this crossing is less likely to be directly involved in a radiationless decay mechanism. Figure 8a shows the energies of all conical intersections, and dotted lines connect the points between which pathways have been found previously.42 The connected points represent the radiationless decay mechanism, and clearly, there is no access to the three-state conical intersection. In adenine, two different three-state degeneracies between the S1, S2, and S3 states have been located at energies close to the vertical excitation energies. These three-state conical intersections are at energies 5.6 and 6.0 eV above the groundstate minimum, respectively. Vertical excitation energies to the two bright states are at 5.7 and 6.1 eV, so the crossings are at similar energies and will probably participate in radiationless pathways that quench fluorescence in adenine. The energies of these points and how they can be connected are shown in Figure 8b. The existence of two different seams of three-state conical intersections indicates that these features are common and complicate the potential energy surfaces of adenine. Furthermore, the presence of three-state conical intersections implies that seams of two-state conical intersections between pairs of these states are branching out of the three-state seam, and these two-state seams are at energies lower than the three-state one. We have calculated points of one of these seams, an S1-S2 seam, that connects the two different three-state conical intersections found here. It has been stated frequently in the past that the dynamics and spectra of excited states in adenine and the other purine bases are affected by the interactions of two excited states. In this work, it is shown that the picture can be a lot more complicated with three or four excited states interacting and crossing. Here, the focus has been on the location of threestate conical intersections. Although the results indicate the potential involvement of three-state conical intersections, further work is needed to verify this involvement. Pathways connecting all the features and any barriers along these pathways are needed. Ultimately, nuclear dynamics can determine the effect of these features in the spectroscopy and dynamics of excited states in adenine. The above results suggest a high frequency of three-state conical intersections in adenine and probably other systems with similar high density of excited states. We anticipate the presence of three-state conical intersections in the other nucleobases. More generally conjugated systems with many excited states and many degrees of freedom are likely to exhibit three-state conical intersections, as well. Acknowledgment. This work was supported by the National Science Foundation under Grant No. CHE-0449853 and Temple University. Supporting Information Available: Tables of internal coordinates for uracil and adenine are given that correspond to geometries optimized and described in the text. This material is available free of charge via the Internet at http://pubs.acs.org. References and Notes (1) Klessinger, M.; Michl, J. Excited States and Photochemistry of Organic Molecules; VCH Publishers: New York, 1995.

Matsika (2) Domcke, W.; Yarkony, D. R.; Ko¨ppel, H. Conical Intersections; World Scientific: Singapore, 2004. (3) Bersuker, I. B. The Jahn-Teller Effect and Vibronic Interactions in Modern Chemistry; Plenum Press: New York, 1984. (4) Katriel, J.; Davidson, E. R. Chem. Phys. Lett. 1980, 76, 259-262. (5) Robb, M. A.; Garavelli, M.; Olivucci, M.; Bernardi, F. ReV. Comput. Chem. 2000, 15, 87. (6) Yarkony, D. R. J. Phys. Chem. A 2001, 105, 6277-6293. (7) Keating, S. P.; Mead, C. A. J. Chem. Phys. 1985, 82, 5102-5117. (8) Han, S.; Yarkony, D. R. J. Chem. Phys. 2003, 119, 11562-11569. (9) Han, S.; Yarkony, D. R. J. Chem. Phys. 2003, 119, 5058-5068. (10) Coe, J. D.; Martinez, T. J. J. Am. Chem. Soc. 2005, 127, 000. (11) von Neumann, J.; Wigner, E. P. Phys. Z. 1929, 30, 467. (12) Matsika, S.; Yarkony, D. R. J. Chem. Phys. 2002, 117, 6907. (13) Matsika, S.; Yarkony, D. R. J. Am. Chem. Soc. 2003, 125, 1067210676. (14) Matsika, S.; Yarkony, D. R. J. Am. Chem. Soc. 2003, 125, 1242812429. (15) Blancafort, L.; Robb, M. A. J. Phys. Chem. A 2004, 108, 1060910614. (16) Kato, S.; Hoenigman, R.; Gianola, A.; Ichino, T.; Bierbaum, V.; Lineberger, W. C. In Molecular Dynamics and Theoretical Chemistry Contractors ReView; Berman, M., Ed.; AFOSR, Shelter Point Hotel: San Diego, 2003; pe 49. (17) Crespo-Hernandez, C. E.; Cohen, B.; Hare, P. M.; Kohler, B. Chem. ReV. 2004, 104, 1977. (18) Daniels, M.; Hauswirth, W. Science 1971, 171, 675. (19) Daniels, M. In Photochemistry and Photobiology of Nucleic Acids; Wang, S. Y., Ed.; Academic Press: New York, 1976; Vol. 1 p 23. (20) Callis, P. R. Annu. ReV. Phys. Chem. 1983, 34, 329. (21) Pecourt, J.-M. L.; Peon, J.; Kohler, B. J. Am. Chem. Soc. 2000, 122, 9348-9349. (22) Pecourt, J.-M. L.; Peon, J.; Kohler, B. J. Am. Chem. Soc. 2001, 123, 10370-10378. (23) Cohen, B.; Hare, P.; Kohler, B. J. Am. Chem. Soc. 2003, 125, 13594-13601. (24) Cohen, B.; Crespo-Hernandez, C. E.; Kohler, B. Faraday Discuss. 2004, 127, 000. (25) Peon, J.; Zewail, A. H. Chem. Phys. Lett. 2001, 348, 255-262. (26) Gustavsson, T.; Sharonov, A.; Markovitsi, D. Chem. Phys. Lett. 2002, 351, 195-200. (27) Gustavsson, T.; Sharonov, A.; Onidas, D.; Markovitsi, D. Chem. Phys. Lett. 2002, 356, 49. (28) Onidas, D.; Markovitsi, D.; Marguet, S.; Sharonov, A.; Gustavsson, T. J. Phys. Chem. B 2002, 106, 11367. (29) Brady, B. B.; Peteanu, L.; Levy, D. H. Chem. Phys. Lett. 1988, 147, 538-543. (30) Nir, E.; Kleinermanns, K.; Grace, L.; de Vries, M. S. J. Phys. Chem. A 2001, 105, 5106-5110. (31) Piuzzi, F.; Mons, M.; Dimicoli, I.; Tardivel, B.; Zhao, Q. Chem. Phys. 2001, 270, 205-214. (32) Lu¨hrs, D. C.; Viallon, J.; Fischer, I. Phys. Chem. Chem. Phys. 2001, 3, 1827. (33) He, Y.; Wu, C.; Kong, W. J. Phys. Chem. A 2003, 107, 51455148. (34) He, Y.; Wu, C.; Kong, W. J. Phys. Chem. A 2004, 108, 943-949. (35) Kang, H.; Lee, K. T.; Jung, B.; Ko, Y. J.; Kim, S. K. J. Am. Chem. Soc. 2002, 124, 12958. (36) Kang, H.; Jung, B.; Kim, S. K. J. Chem. Phys. 2003, 118, 6717. (37) Kim, N. J.; Jeong, G.; Kim, Y. S.; Sung, J.; Kim, S. K.; Park, Y. D. J. Chem. Phys. 2000, 113, 10051. (38) Ullrich, S.; Schultz, T.; Zgierski, M. Z.; Stolow, A. J. Am. Chem. Soc. 2004, 126, 2262. (39) Ullrich, S.; Schultz, T.; Zgierski, M. Z.; Stolow, A. Phys. Chem. Chem. Phys. 2004, 6, 2796-2801. (40) Ismail, N.; Blancafort, L.; Olivucci, M.; Kohler, B.; Robb, M. A. J. Am. Chem. Soc. 2002, 124, 6818-6819. (41) Mercha´n, M.; Serrano-Andre´s, L. J. Am. Chem. Soc. 2003, 125, 8108-8109. (42) Matsika, S. J. Phys. Chem. A 2004, 108, 7584-7590. (43) Sobolewski, A. L.; Domcke, W. Eur. Phys. J. D 2002, 20, 369374. (44) Perun, S.; Sobolewski, A. L.; Domcke, W. J. Am. Chem. Soc. 2005, 127, 6257-6265. (45) Dunning, T. H. J. Chem. Phys. 1970, 53, 2823. (46) Lischka, H.; Shepard, R.; Pitzer, R. M.; Shavitt, I.; Dallos, M.; Myller, T.; Szalay, P. G.; Seth, M.; Kedziora, G. S.; Yabushita, S.; Zhang, Z. Phys. Chem. Chem. Phys. 2001, 3, 664. (47) Lischka, H.; Dallos, M.; Szalay, P. G.; Yarkony, D. R.; Shepard, R. J. Chem. Phys. 2004, 120, 7322. (48) Dallos, M.; Lischka, H.; Shepard, R.; Yarkony, D. R.; Szalay, P. G. J. Chem. Phys. 2004, 120, 7330. (49) Lischka, H.; Dallos, M.; Shepard, R. Mol. Phys. 2002, 100, 1647.

Three-State Conical Intersections in Nucleic Acid Bases (50) Broo, A.; Holme´n. Chem. Phys. 1996, 211, 147. (51) Broo, A. J. Phys. Chem. A 1998, 102, 526. (52) Mennucci, B.; Toniolo, A.; Tomasi, J. J. Phys. Chem. A 2001, 105, 4749-4757. (53) Jean, J. M.; Hall, K. B. J. Phys. Chem. A 2000, 104, 1937. (54) Salter, L. M.; Chaban, G. M. J. Phys. Chem. A 2002, 106, 4251. (55) Mishra, S. K.; Shukla, M. K.; Mishra, P. C. Spectrochim. Acta, Part A 2000, 56, 1355.

J. Phys. Chem. A, Vol. 109, No. 33, 2005 7545 (56) Fu¨lscher, M. P.; Serrano-Andres, L.; Roos, B. O. J. Am. Chem. Soc. 1997, 119, 6167-6176. (57) Clark, L. B.; Peschel, G. G.; Tinoco, I., Jr. J. Phys. Chem. 1965, 69, 3615. (58) Atchity, G. J.; Xantheas, S. S.; Ruedenberg, K. J. Chem. Phys. 1991, 95, 1862. (59) Schaftenaar, G.; Noordik, J. J. Comput.-Aided Mol. Design 2000, 14, 123-134.